Soft-Computing Based Trust Management …Soft-Computing Based Trust Management Framework for Group...
Transcript of Soft-Computing Based Trust Management …Soft-Computing Based Trust Management Framework for Group...
Received: March 22, 2017 327
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
Soft-Computing Based Trust Management Framework for Group Key
Management in MANETs
Nagaraja Gundluru1* Pradeep Reddy Ch1
1School of Information Technology & Engineering, VIT University, Vellore, India
* Corresponding author’s Email: [email protected]
Abstract: We can realize instant joint group communication by forming Mobile Ad Hoc Networks without
demanding any pre-plan or pre-existing infrastructure setup. Conversely, the curbs of these networks such as
unreliable wireless medium, unpredictable topology, no central administration, fuel the compulsion of a key based
cryptographic algorithm to defend data traffic. In this perspective, substantial research work has been done in the last
decade or so and ascertained that the trust based frameworks for group key management deliver superior
performance than others. Since the nodes in ad hoc networks have limited computing resources, the overall
performance of the system depends on how effectively and securely designed the system. This encourages us to
work on a framework which consumes less computing power and also invulnerable to internal as well as external
attacks. We propose a framework which reduces the network resource consumption for the trust request and
collection using game theory concept. The energy of the wireless nodes are significantly saved by choosing the novel
strategy called as finding optimal set of remote nodes to send the response for trust request using game theory.
Choosing local optimal at each stage eventually leads to global optimal. Later, synthesize the collected trust and
handle the attacks using fuzzy concept in order to get the degree of trustworthiness instead of binary classification.
We prove with our simulation results that our proposed scheme reduces overhead of the network significantly while
without compromising security aspects.
Keywords: Mobile ad hoc network, Group key management, Game theory, Fuzzy logic, Trust management
framework, Attacks, Dishonest nodes.
1. Introduction
1.1 Mobile ad hoc networks
A mobile ad hoc network (or MANET) is a
wireless network comprises of mobile nodes which
need petty or infrastructure-less to deploy instantly
and facilitate group communication. It has a
dynamic topology due to a node may join in, leave
from, or move around the network at any point of
time [1]. Since a MANET can be rapidly and
suddenly organized, it has strengthened
attractiveness in cooperative application situations
such as disaster rescue operations, battlefields,
conferences, etc. The majority of these setups
assume there would be an efficient and secure group
communication framework [2]. The hurdles to build
such framework in MANETs comprise constrained
computing power (i.e. Bandwidth, Battery, CPU,
Memory, etc.) of each wireless node, untrustworthy
wireless medium and irregularity in network
topology due to node mobility. In a MANET, there
is a direct communication between neighbors within
the range of wireless medium; otherwise, via
intermediate nodes if nodes are out of range. Each
node acts as a terminal which sends or receives data
and also router in order to cooperate for
communication of other nodes.
1.2 Role of group key management
Group key can be considered as a Traffic
Encryption Key (TEK) which plays a vital role in
secure group communication systems and have a
couple of important components called as security
Received: March 22, 2017 328
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
and efficiency [3, 4]. The security module
safeguards group member authentication, group
message's integrity and confidentiality, node
compromise robustness, forward and backward
secrecy, immediate rekeying, and group
independence. The efficiency module safeguards
scalability, flexibility, low storage, low computation
and low communication overhead.
1.3 Trust management framework (TMF)
The “trust” concept was primarily presented by
social sciences and is defined as the degree of
subjective belief about the behaviors of a particular
entity [5]. In the context of networking, it can be
considered as the belief of an assessing node about
the truthful nature of assessed node based on the
experience got from past interactions. The trust
characteristics are dynamic, asymmetric, context
dependent, not transitive and subjective. In fact, the
MANET concept works based on the cooperation
among the nodes. However, due to some node
exhibits selfish (i.e. save its resources without
cooperating to other nodes) and malicious (i.e.
populate fake routes, deny network service, drop
packets, etc.) behavior leads to the necessity of a
trust management framework (TMF) [6]. The
purpose of TMF is to boost the collaboration in the
network while penalize selfish or malicious behavior
nodes. It consists of four modules, namely, trust
request, trust collection, synthesize collected trust
and apply the result for applications such as routing,
key management, resource management, etc. The
trust information collection module gathers the
information about nodes’ behavior from local (i.e.
neighbors) nodes and recommendations from remote
nodes upon sending requests [7]. The trust
synthesizes module evaluates trustworthiness (i.e.
how much a node can believe in other nodes) of
each node based on collected information. The trust
application module deduces if a node can be trusted
based on its trustworthiness level.
1.4 Attacks on TMFs
We can classify the attacks posed by malicious
nodes in the network into two, namely, passive and
active attacks. With passive attacks, they can just
copy the data traffic while with active they can
modify the traffic also. To benefit from a system
failure, selfish or malicious nodes can even send
biased recommendations through attacks such as
Black whole attack, Denial of Service attack, On-
Off attack, Lying attack, Selective attack, Positional
and Seasonable attacks [8, 9].
1.5 Game theory and fuzzy logic
Game theory is the concept of applied
mathematics which provides mathematical tools to
analyze the outcome of sequence complex decisions
taken by several rational entities [10]. In the context
of MANETs, it can be formally defined as a 3-tuple
game, G = <N, A, {ri}>, where N = {1, 2, ..., N} is
the set of wireless nodes in the network, A is the
possible strategy chosen by each node, and for all
players it is the Cartesian product, A = A1 x A2 x ...
x An, and finally ri is the reputation or payoff
function {ri} = {r1, r2, … ,rn} which can get by a
node if the particular strategy followed. So each
node tries to maximize its reputation by choosing an
optimal strategy. One of the solutions suitable in the
context of MANETs for game theory is called as
Nash equilibrium [11-12]. It gives a chance to get
the optimal gain to all wireless nodes. No node
should take its own strategy to benefit more payoffs
while other node’s payoffs are not improved. The
Nash equilibrium is the best solution which
comprises a set of strategies that nodes can choose
and get corresponding payoffs.
Fuzzy logic is the concept used to handle
uncertainty situations in automated artificial
intelligence applications which have imprecise
information and need to take decisions [13]. The
result will be “degree of truth” rather than usual
binary values “true” or “false”. For example, the
predicate, Today is sunny, might be 100% true if
there are no clouds, 75% true if there is a slight
cloud, 50% true if it is cloudy and 0% true if it
showers all day. So instead of taking a binary
decision, we will consider the options in between
with this concept.
1.6 Problem Identification
There is a necessity to protect the data traffic
from the internal attacks such as misbehavior or
selfish node and external attacks by other
competitor’s due to curbs of the MANETs such as
wireless medium, no central administration. In this
context, most of the cryptographic algorithms which
relies on group key such as symmetric and
asymmetric algorithms proposed. But, these
techniques involve significant computing resources
and energy consumption. Moreover, these expect a
central administrator, but which is lacking in
MANETs. So the alternative technique proposed in
the literature is called as trust based framework
which significantly reduces the overhead of
computation, energy required, and also resistance to
attacks. Here, trust plays a vital role and the process
of trust request, collect, synthesize and apply needs
Received: March 22, 2017 329
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
significant bandwidth and energy of wireless nodes
in the network. So we need a system which performs
the same with less computational resources of the
network without compromising security. This
motivates us to work on this paper. Our proposing
framework performs local optimization at each stage
so that it leads to the global optimum. As far as our
knowledge, no specific work considered an optimal
solution at all stages like we do here. The
contribution of this paperwork is to design a
strategic trust management framework which would
be
1. Send trust request to optimal number of nodes
so that saves bandwidth and energy.
2. Trust collects from the optimal number of
nodes using game theory concept.
3. Synthesize the collected trust using fuzzy logic.
4. Use irregularity pattern strategy to handle
dishonest remote nodes in the network.
The rest of this paper is arranged as follows. In
Section 2, the trust management frameworks
proposed so far for MANETs briefed and the
motivation for our work also discussed. We present
our proposed optimized trust management
framework in Section 3. In Section 4, we elucidate
the simulation results of our proposed framework in
terms of how the overhead of the network
significantly reduced while maintaining security.
The section 5 provides conclusions and future work
aspects.
2. Related work and motivation
In the last decade or so, several trust
computation models and trust based key
management frameworks have been proposed for
MANETs. However, the attention paid about
dealing with energy saving of the wireless nodes is
not up to the mark. In this section, we present the
proposed models for key management and discuss
their pros and cons.
In [13], the authors proposed a mechanism
called as Reliable Group Key Management
Framework using Fuzzy Logic for MANETs
(RGMFFL). This is to detect and exclude the
malicious nodes from further communication using
the fuzzification and defuzzification process instead
of binary classification. However, they didn’t
address the overhead related to the trust computation
process. In [14], the authors proposed a self-
organizing trust based security architecture for key
management in MANETs. It works by establishing
keys between nodes based on their trust level and
trust relationships. The advantage of this approach is
that it considers the trust as physical as well as a
logical entity. However, establishing pairwise keys
based on trust may not be realistic in the context of
MANETS due to high scalability and network
dynamics. In [15], the authors proposed a
hierarchical key management framework which
adopts Public Key Infrastructure (PKI) model where
nodes can dynamically take management roles. It
offers redundancy and robustness in the formation
of Security Association (SA) between pairs of nodes.
However, the certificate chains are used to derive
trust relationships. In [16], the authors suggested a
hop-by-hop and on-demand public key management
protocol for MANETs. Here, each node makes its
own public/private key pairs, issues its certificate to
neighboring nodes, preserves received certificates in
its certificate repository, and provides authentication
service by adjusting to the dynamic network
topology, without depending on a centralized server.
However, the certificate chains are used to derive
trust relationships.
In [17], the authors proposed a trust model based
on Markov chain to get the trust values for 1-hop
neighbors. They designed a trust-based hierarchical
key management scheme by selecting a certificate
authority server (CA) and a backup CA with the
highest trust values. This work contributes a severe
analysis of trust values and studies a range of attacks.
However, it calculates trust, only based on direct
interactions and does not consider indirect trust
recommendations from remote nodes. In [18], the
authors proposed a survey of key management
techniques targeted to only network-layer security.
In [19], the authors proposed a framework to
mitigate double-face attacks based on collecting
both direct and remote recommendations. However,
it is not resistant to bad mouthing and iterative on-
off behaviors. Here, the trust is assessed based on
traffic via neighbor node and so it is time consuming.
In [20], the authors proposed a protocol
independent and self-adaptive scheme named
Autonomic Trust Knowledge Monitoring Scheme
(ATMS). It uses autonomic management model to
optimize resource consumption. However, ATMS is
vulnerable to lying attacks due to there is no
mechanism to separate genuine and fake trust
recommendations. It is also not resistant to on-off
attack due to not maintain a history of nodes. In [21],
the authors proposed a multipath routing protocol
for MANETs to encounter double-face attacks.
However, it is vulnerable to positional attack due to
recommendations are not broadcasted across the
network and also weak to lying attacks. In [22], the
authors proposed a dynamic nature-inspired model
which calculates the trust level of nodes based on
general data classes. However, the framework
Received: March 22, 2017 330
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
cannot survive with the on-off attack and regional
attacks.
In [23], the authors proposed an encryption
based framework by extending AODV [24] protocol
named as Trusted AODV. They used consensus
algorithm to resist conflicting behavior attack.
However, this framework leads to time consuming
process due to the reactive nature of trust
recommendations. In [25], the authors proposed
trust management framework based on fuzzy logic.
It combines the node’s serving capability (i.e.,
bandwidth, remnant battery, CPU, memory, etc.)
and behavior in order to compute trust. However,
the performance of this approach depends on trust
information collection method which is not defined
clearly. It is also vulnerable to on-off attack due to
not distinguishing honest and fake behaviors. In
[26], the authors proposed a framework to deal with
lying and double-face attacks. It evaluates the
trustworthiness of a node from diverse angles such
as context, severity of the outcome, etc. However,
the network leads to instability due to not providing
a uniform view of trust values across the network. It
is immune to lying attacks, but depends on the
accuracy of the methods used to evaluate the
recommendations.
In [27], the authors proposed a trust-based
extension of AOMDV [28] (a multi-path extension
of AODV), named as Ad hoc On-demand Trusted-
path Distance Vector (AOTDV) to resist bad
mouthing and double-face attacks. One important
observation here is that it considers the data and
control packets separately. Though it is resistant to
some attacks, the black list feature is a very time
consuming process. A defence scheme for the
recommendation based trust model is proposed in
order to handle some attacks posed by dishonest
recommendations, named as Cluster Based
Recommendation Filtering (CBRF) [29]. It uses the
clustering technique based the level of confidence,
deviation threshold, and closeness centrality value to
ensure that recommending node is a close friend to
the evaluating node for a period of time. However,
this model is heavyweight and it consumes more
computing power of nodes.
In [30], the authors proposed a risk strategy
model to determine the optimal number of
recommendations that can satisfy the security
requirements of a network. Based on the optimal
number of recommendations, they introduced a new
trust derivation scheme. The probability of the
selected strategy was calculated based on the mixed
strategy Nash equilibrium of the game. Compared
with the traditional trust derivation methods, the
simulation results showed that this game theoretic
approach can improve the performance of the
network under the premise of security assurance,
especially in a dense network. But, they didn't
consider the same concept for trust request, which
can further improve the performance of the network.
What follows from aforesaid discussion is that
the contributions made from several researchers so
far mostly related to the trust computation models
and mitigating the attacks on the same. A little bit
work happened related to saving the computing
power of a wireless node in the context of MANETs.
So, we considered this issue and proposing a
framework which consume less energy of the node
while without compromising security..
3. Proposed framework
In this paper work we consider the MANET
which consists of a set of wireless nodes with
dynamic topology and there is no central
administrator to monitor and control the network.
Here, we adopt the work proposed in [13] and
extending the same to save energy for trust
computation process using soft computing technique
called as game theory concept [STGM]. The
network will be virtually clustered [31] just to
organize the hierarchy among nodes and then the
direct communication happen between nodes within
the cluster, but the communication between nodes
from two different clusters will be taking place via
cluster leader’s as shown in Fig.1. It consists of
three clusters C1, C2 and C3. Here, N3, N7, and N11
are chosen as cluster leaders CL1, CL2, and CL3
respectively based on their highest trust value,
remaining energy and low mobility.
We assume that each wireless node has a data
structure as in Table 1 and Table 2 to maintain
dynamic essential information such as who are the
neighbor nodes, own trust value, trust
recommendation on neighbor nodes, remaining
energy and its mobility rate. This information can be
periodically updated through flooding. We quantify
the trust as between -1 and +1. The -1 indicates that
a node is completely malicious or dishonest and +1
indicates that a node is completely genuine or honest.
We consider the trust value as 0.5 for a newly joined
node. The trust value of a node is figured as the
mixture of direct and indirect (by remote nodes
recommendations) observations as shown in Fig. 2.
Here, the node X is evaluating node and Y is
evaluated node. The following Eq. (1) is used to
compute the trust value of a node.
Trust (NX, NY) = tanh (TDirect + TRecommend) (1)
Received: March 22, 2017 331
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
TDirect = (∑ 𝑊𝑖𝐷𝑖𝑛𝑖=1 ) (2)
TRecommend =∑ 𝑇𝑅𝑗 (𝑇𝑅(𝑁𝑗 , 𝑁𝑌))𝑚𝑗=0 (3)
Where,
n = The no.of direct observations between X and Y.
Di = the value between -1 and +1 based on the direct
observation is bad or good respectively.
Wi = Weight for each direct observation based on
importance.
m = The no.of remote nodes giving trust
recommendations (indirect trust) on Y.
TRj = Trust value of remote node j who sends
indirect trust.
TR(Nj, NY) = Trust value on Node Y by its
neighbor’s node j.
The use of the hyperbolic tangent function is to
make the trust value between -1 and +1 despite of
its value out of bounds. Here the remote nodes are
R1, R2, and R3 whose have direct observation with
node Y. So, the node X collects trust from these
remote nodes and aggregated.
We assume that there are an M number of
neighbor nodes for the evaluated node. Let us
consider first how the trust request process is to be
optimized. Send the trust request to set of remote or
neighbor nodes of the evaluated node whose trust
value is above the threshold (Here, considered as
0.5), more remaining energy (Here, considered as
5J) and low mobility rate (Here, considered as less
than 5m/Sec). This procedure is to filter the number
of remote nodes to whom the trust request sends. It
saves the bandwidth of the network as well as saves
energy for forwarding routing packets.
Table 1. A data structure to hold basic information
about wireless nodes in network
Node
(Ni)
Neighbours
(Nj)
Trust
(Ni)
Direct Trust
on Nj by Ni
A B, C, E 0.8
(B, A) = 0.5
(C, A) = -0.7
(E, A) = 0.2
Table 2. A data structure to hold additional
information about wireless nodes in network
Node
(Ni)
Remaining
Energy (Ni )
Mobility
Rate (Ni)
A 11 J 10 m/sec
Figure.1 Formation of Clusters
Figure.2 Computing Direct and Indirect Trust
To get the response from an optimal set of nodes
the concept used in [30] is considered. We believe
that the network is secure if and only if we consider
all the recommendations trust into account. But, by
taking the security features of the network into
consideration, we define a payoff U under the
condition that the number of recommending nodes
that respond to the evaluating node is k. To simplify
the analysis, we assume that the energy consumed
by a remote node to send trust response is Ns(e) and
assume that the corresponding gain from this task is
appropriate. Consequently, we have U > kNs(e) > 0.
Assume that the requirements of security can be
fulfilled if any node sent trust response. The
procedure for strategies of nodes and their payoff
Received: March 22, 2017 332
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
matrix can be formulated in Table 3. The row player
(node i) can be a random contributing node that
receives the trust request, while the column player
stands for the other k-1 neighbor nodes.
Here, we can make a node to stop from sending
the response to save energy based on the assumption
that other nodes can trust honest response and they
make the network work. As a random node chooses
its own strategy, all the wireless nodes are
independent in this game. We assume that a random
node i send a trust response with probability p, or
remains stopped with probability 1-p. Then out of k
nodes, at least one node replies the trust request with
a probability of 1-(1-p) k. As a result, the mixed
strategy Nash equilibrium can be computed by Eq.
(4) and consequently, the probability of sending
trust reply p can be computed and then q = (1-p)
will be computed.
U (1-(1-p) k-1) = U - Ns(e) (4)
Once the direct and indirect trust gathered together,
the final mixture will be synthesized further using
fuzzy logic instead of taking a binary decision. The
Table 4 describes about the working model of
synthesizing module in the trust management
framework using a fuzzy logic system with Rule
base (i.e. IF-THEN control structures) to eliminate
recommendations from dishonest nodes from the
network to avoid further communication with those
nodes.
Table 3 Payoff Matrix for any one node sends a
response (k=1)
For (k=1)
Other Remote Nodes (M -1)
No Other
Response
At Least One
Response
(1 ≤ β ≤ M – 1)
Node
Ni
No
Response
0,0 U, U – β Ns(e)
Response U – Ns(e) ,
U
U – Ns(e) , U – (β
Ns(e))
Table 4 Fuzzy Set Membership Function
Trust value
of a node
Trustworthiness
of a node
Recommendation
risk of a node
0.5 to +1 Excellent NIL
0 to 0.49 Very Good Low
-0.6 to -0.99 Good Moderate
-0.3 to -0.59 Fair High
-1 to -0.29 Poor Very High
The pattern irregularity in the recommendations
will be checked in order to handle the on-off attack
in which the malicious nodes behavior will switch
between normal (honest recommendations) and
abnormal (dishonest recommendations) over a span
of time. We maintain the history of the
recommendations made by remote assessing nodes
on a particular assessed node, didn’t consider the
recommendations which are too far from the
variance and then taking into account the average of
all that. So, it will not affect much on the trust value
of a node in particular duration. The proposed
optimal trust management framework follows the
below steps periodically.
1. Initialize the Network.
2. if Node Ni is ready then
3. flood the basic information to all nodes
in the Network.
4. end if
5. Form the Clusters and Select the Cluster
Leader’s.
6. One of the Cluster Leader CLi initiates the
trust computation periodically.
7. To collect indirect trust for node Nj by Ni
8. Ni sends trust request to only set of
neighbor nodes (M) whose trust value is
above threshold (Trust (Ni) >
Threshold), high remaining energy and
Low Mobility.
9. Select optimal k < M, which is sufficient to
recommend trust instead of all neighbor’s.
10. Compute the final trust based on Direct and
Indirect trust.
11. Synthesize the calculated trust value based
on following Fuzzy Rules:
12. If (Trustworthiness(Ni) is Excellent)
Then Recommendation_Risk(Ni) is NIL
13. If (Trustworthiness(Ni) is Very Good)
Then Recommendation_Risk(Ni) is Low
14. If (Trustworthiness(Ni) is Good) Then
Recommendation_Risk(Ni) is Moderate
15. If (Trustworthiness(Ni) is Fair) Then
Recommendation_Risk(Ni) is High
16. If (Trustworthiness(Ni) is Poor) Then
Recommendation_Risk(Ni) is Very
High
17. Select the Cluster Leader based on latest
trust values.
18. Repeat step 2 to 18 periodically.
4. Simulation and results
4.1 Simulation settings
To simulate our proposed framework, we used
NS2 simulator [32] tool which is an open source
discrete event simulator exclusively designed to
promote research in the field of computer networks
Received: March 22, 2017 333
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
including MANETs. We used the Multicast AODV
routing protocol in order to benefit from
multicasting feature. The network is simulated in the
area of 1000 X 1000 square meters with 100 mobile
nodes with simulation time is 500Secs. The network
configuration settings are shown in Table 5. In order
to test the network performance with the proposed
trust management framework, we considered
metrics such as Packet Delivery Ratio, Packet Loss,
Residual Energy, Detection Ration of malicious
nodes, and Group Key Management Overhead in the
cases of our proposing work, STGM and existing
framework named as RGMFFL [13]. The results are
plotted from Fig. 3 to Fig. 7. We assumed that there
are maximum 60% of malicious nodes (dishonest
recommending nodes) in the network and the
percentage of malicious nodes is increased to the
maximum level during the span of simulation time,
500Secs. It is witnessed that the network packet
delivery ratio with STGM at in the range of 88% to
70%, while with RGMFFL framework the same
falls from 88% to 28%. This improvement is
because of the elimination of malicious nodes by
discriminating them by degree of trustworthiness of
the nodes predicted through fuzzy classification.
The fuzzy classifier helps to find the group of lowest
trustworthy nodes with highest elimination risk.
Here, the detection ratio of malicious nodes has
been increased over the simulation time and so there
is a significant increase in the packet delivery ratio
and decrease in packet loss. As we are also
maintaining the history of recommendations given
by a remote node on a particular node, the mean
value of the trust will be considered as final. It helps
to avoid the too far away recommendations from the
remote nodes.
Table 5. Simulation Settings
Number of Nodes 100
Area 1000 X 1000 square meters
MAC 802.11
Simulation Time 500 Sec
Traffic Source CBR
Mobility Model Random Waypoint
Speed 10 m/Sec
Pause Time 50 Sec
Routing Protocol MAODV
Initial Energy 15 J
Trust Threshold 0.3
Radio Range 200 m
Propagation Two-ray ground reflection
model
Figure.3 Malicious nodes Vs packet delivery ratio
Figure.4 Malicious nodes Vs packet loss
Figure.5 Malicious nodes Vs average residual energy
Figure.6 Malicious nodes Vs detection ratio of malicious
nodes
Received: March 22, 2017 334
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
Figure.7 Malicious nodes Vs key management overhead
It is also observed that the percentage of packet
loss without optimized framework upsurges while
reduced significantly with optimized framework. To
simulate the residual energy of nodes we considered
the average of all the nodes in the network. The
behavior of energy consumed by wireless nodes in
the MANET has fallen from 14J to almost 8J with
our proposed framework, STGM and while the same
fallen from 14J to nearly 2J with the existing
RGMFFL framework. The energy saved due to send
the trust request itself to a subset of the nodes
among all neighbors without compromising the
security of the network. This is because with the
help of game theory concept called as Nash
equilibrium; we will get the optimal (necessary and
sufficient) subset of the nodes that will send the
response to the trust request from assessing node on
assessed node. So the bandwidth of the network
saved as well as the energy of the remaining nodes
will be saved without receiving requests and
responding to the same.
The detection ratio of malicious nodes is in the
range of between 100% and 70% with our
framework, STGM and while the same is between
100% and 30% with the existing RGMFFL
framework. This is because of the fuzzy classifier
which will get the degree of genuineness and do the
classification between honest and dishonest nodes.
Further the rate of detection of malicious nodes is
decreased due to the remote nodes are colluding
together to avoid from the detection by fuzzy
classifier. The key management overhead is
observed as the number of control packets used.
Almost 40% of significant key management
overhead is reduced with our proposed framework
while comparing to the existing framework. There
are three reasons for the significant reduction in the
key management overhead. The first one is that we
are sending the request to the nodes whose trust
value is excellent and has more energy. The second
one is to get the response from a subset of the nodes
only. So the number of control packets will be saved
and so the overhead of the network will be saved for
each time of the re-keying process due to a new
node joins into the network or an existing node left
of the network.
4.2 Security threat model
For testing the proposed security framework, we
have launched outsider attacks such as replay attack,
node capture attack, data manipulation attack and
selective forwarding attacks. This approach
considers the insider attacks such as the Packet
dropping attack, false trust report attack, report
disruption attack, false join or leave requests, battery
exhaustion attack (DDoS). Because of using key
management and authentication, outsider attacks are
avoided from the network. By using the trust
mechanism, insider attacks are also detected and
prevented. To simulate about the handling of
internal attacks, let us consider the network with 20
nodes with 3 clusters virtually. Assume the
corresponding trust values of nodes and their
clusters are as shown in the Table 6.
Let us consider the source S = Node 3 and
destination D = Node 20. Let P1, P2 and P3 be the
possible routes determined for S to D given by
P1 => 3-2-9-11-14-17-20
P2 => 3-2-7-11-12-14-20
P3 => 3-2-9-11-10-16-17-20
Let TRP1, TRP2 and TRP3 be the total global
trust values of the paths P1, P2 and P3, respectively,
given by
Table 6. Global Trust values of all nodes
Node
id
Cluster
Id
Global Trust
value (TR)
1 C1 0.40
2 C1 0.70
3 C1 0.54
4 C1 0.64
7 C1 0.33
8 C2 0.74
9 C2 0.42
10 C2 0.28
11 C2 0.81
12 C2 0.23
13 C2 0.69
14 C3 0.52
15 C3 0.45
16 C3 0.75
17 C3 0.81
18 C3 0.48
20 C3 0.65
Received: March 22, 2017 335
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
TRP1 =
0.54+0.70+0.42+0.81+0.52+0.81+0.65 =4.45
TRP2 =
0.54+0.70+0.33+0.81+0.23+0.52+0.65 =3.78
TRP3 =
0.54+0.70+0.42+0.81+0.28+0.75+0.81+0.65
=4.15
Hence the path P1 which is having the highest
global trust value (4.45) is selected, whereas the
paths P2 with least trust value (3.78) and P3 with
less trust value (4.15) are not selected, thereby
omitting the internal attacker’s nodes 7, 10 and 12.
5. Conclusion and future works
In this paper, we proposed a soft-computing
based trust management framework for handling the
process of group key management in the MANETs.
Here trust value is determined for each node based
on the direct and indirect observations. In this
context, the trust request and collection plays a vital
role and so leads to the overhead of the network. To
minimize the same we applied the concept of game
theory to send the trust request itself to the subset of
the remote nodes whose responses are necessary and
sufficient without affecting the security of the
members in the network. The concept of Nash
equilibrium gives the best strategy to choose at
every time of trust calculation and so saving the
bandwidth of the network and energy of the nodes.
Later the trust value of all nodes would be
interpreted with fuzzy classifier in order to find the
group of lowest trustworthy nodes whose
elimination risk at very high. Those nodes will be
excluded from further communication in the
network. Then, the network is clustered and the
cluster leader is elected based on the highest trust
vale, highest remaining energy and lowest mobility
rate. The working mechanism of the proposed
framework described through an algorithm. By
simulation results, we proved that the proposed
technique reduces the number of control packets
required to manage a group key and so leads to save
energy of each node during the re-keying process. A
weakness of using fuzzy logic is that storing the
rules database might involve a significant amount of
memory. It is also good if it has a mechanism to
evade from isolating remote alone nodes. As a
future work direction, we plan to investigate
aforementioned issues and defence mechanisms for
attacks such as bad-mouthing, ballot-stuffing, and
collusion posed by dishonest nodes.
References
[1] K.K. Chauhan and A.K. Singh Sanger, “Securing
mobile ad hoc networks: key management and
routing”, International Journal on Ad Hoc
Networking Systems, Vol. 2, No. 2, 2012.
[2] S. Rafaeli S and D. Hutchison D, “A survey of key
management for secure group communication”, ACM
Computing Surveys, Vol. 35, No. 3, pp. 309–329,
2003.
[3] O. Cheikhrouhou, “Secure group communication in
wireless sensor networks: a survey”, Journal of
Network and Computer Applications, Vol. 61, pp.
115-132, 2016.
[4] G Nagaraja and Pradeep Reddy CH, “A survey on
group key management frameworks for secure group
communication in mobile ad hoc networks”,
International Journal of Pharmacy & Technology,
Vol. 8, pp. 4121-4129, 2016.
[5] Karen S. Cook, “Trust in society,” 1st ed. New York:
Russell Sage Foundation, 2001.
[6] J.-H. Cho, A. Swami, and I.-R. Chen, “A survey on
trust management for mobile ad hoc networks”, IEEE
Communications Surveys & Tutorials, Vol. 13, pp.
562–583, 2011.
[7] X. Li, J. Slay, and S. Yu, “Evaluating trust in mobile
ad hoc networks,” In: Proc. of the International
Conference on Computational Intelligence and
Security, pp. 1-10, 2005.
[8] Z. Movahedi, Z. Hosseini, F. Bayan, and G. Pujolle,
"Trust-distortion resistant trust management
frameworks on mobile ad hoc networks: a survey",
IEEE Communications Surveys & Tutorials, Vol. 18,
pp. 1287-1309, 2016.
[9] G. Nagaraja and Pradeep Reddy CH, “Mitigate lying
and on-off attacks on trust based group key
management frameworks in MANETs”,
International Journal of Intelligent Engineering and
Systems, Vol. 9, No.4, pp. 215-224, 2016.
[10] C. Kamhoua, N. Pissinou, and K. Makki, “Game
theoretic modeling and evolution of trust in
autonomous multi-hop networks: application to
network security and privacy”, In: Proc. of the IEEE
International Conference on Communications, pp. 1–
6, 2011.
[11] M. Naserian and K. Tepe, “Game theoretic approach
in routing protocol for wireless ad hoc networks”, Ad
Hoc Netw., Vol. 7, No. 3, pp. 569–578, 2009.
[12] L. Dasilva, H. Bogucka and A. Mackenzie, "Game
theory in wireless networks", IEEE Communications
Magazine, Vol.49, No.8, pp.110-111, 2011.
[13] G. Nagaraja and Pradeep Reddy CH, “A Reliable
Group Key Management Framework Using Fuzzy
Logic for MANETs”, International Journal of
Intelligent Engineering and Systems, Vol.9, No.4,
pp.107-115, 2016.
[14] M. Virendra, M. Jadliwala, M. Chandrasekaran, and
S. Upadhyaya, “Quantifying Trust in Mobile Ad-Hoc
Networks”, In: Proc. of the Int’l Conf. Integration of
Received: March 22, 2017 336
International Journal of Intelligent Engineering and Systems, Vol.10, No.3, 2017 DOI: 10.22266/ijies2017.0630.37
Knowledge Intensive Multi-Agent Systems, pp. 65-70,
2005.
[15] G. C. Hadjichristofi, W. J. Adams, and N. J. Davis,
“A Framework for Key Management in a Mobile Ad
Hoc Network”, In: Proc. of the Int’l Conf. on
Information Technology: Coding and Computing,
Tiejun Huang, Vol. 2, pp. 568-573, 2005.
[16] R. Li, J. Li, P. Liu, and H. H. Chen, “On Demand
Public Key Management for Mobile Ad Hoc
Networks”, Wiley’s Wireless Communications and
Mobile Computing, Vol. 6, No. 3, pp. 295-306, 2006.
[17] B. J. Chang and S. L. Kuo, “Markov Chain Trust
Model for Trust Value Analysis and Key
Management in Distributed Multicast MANETs”,
IEEE Transactions on Vehicular Technology, Vol. 58,
No. 4, pp. 1846-1863, 2009.
[18] A. Hegland, E. Winjum, S. F. Mjolsnes, C. Rong, O.
Kure and P. Spilling, “A Survey of Key Management
in Ad Hoc Networks”, IEEE Commun. Surveys and
Tutorials, Vol. 8, No. 3, pp. 48-66, 2006.
[19] G. Bella, G. Costantino, and S. Riccobene,
“Managing reputation over MANETS,” In: Proc. of
the 4th Int. Conf. Inf. Assur. Security, pp. 255–260,
2008.
[20] Z. Movahedi, M. Nogueira, and G. Pujolle, “An
autonomic knowledge monitoring scheme for trust
management on mobile ad hoc networks”, In: Proc.
of the IEEE Wireless Communications and Networks
Conf., pp. 1898–1903, 2012.
[21] S. Almotiri and I. Awan, “Trust routing in MANET
for securing DSR routing protocol,” PGNet, 2010.
[22] M. Seredynski and P. Bouvry, “Nature-inspired
evaluation of data classes for trust management in
MANETS”, In: Proc. of the IEEE 26th Int. Parallel
Distributed Processing, Symposium, Workshops &
PhD Forum, pp. 366–373, 2011.
[23] X. Li, M. R. Lyu, and J. Liu, “A trust model based
routing protocol for secure ad hoc networks,” In:
Proc. of the IEEE Aerosp. Conf., Vol. 2, pp. 1286–
1295, 2004.
[24] C. E. Perkins and E. M. Royer, “Ad-hoc on-demand
distance vector routing”, In: Proc. of the 2nd IEEE
Workshop Mobile Comput. Syst. Appl., pp. 90–100,
1977.
[25] H. Xia, Z. Jia, L. Ju, X. Li, and Y. Zhu, “A
subjective trust management model with multiple
decision factors for MANET based on AHP and
fuzzy logic rules”, In: Proc. of the IEEE/ACM Int.
Conf. Green Computing Communications, pp. 124–
130, 2011.
[26] W. Li, A. Joshi, and T. Finin, “Coping with node
misbehaviors in ad hoc networks: A multi-
dimensional trust management approach”, In: Proc.
of the 11th Int. Conf. Mobile Data Manage., pp. 85–
94, 2010.
[27] X. Li, Z. Jia, L. Wang, and H. Wang, “Trust-based
on-demand multipath routing in mobile ad hoc
networks,” IET Inf. Security, Vol. 4, No. 4, pp. 212–
232, 2010.
[28] Y. Yuan, H. Chen, and M. Jia, “An optimized ad-
hoc on-demand multipath distance vector (AOMDV)
routing protocol”, In: Proc. of the Asia-Pac. Conf.
Commun., pp. 569–573, 2005.
[29] A.M. Shabut, K.P. Dahal, S.K Bista, and I.U. Awan,
"Recommendation Based Trust Model with an
Effective Defence Scheme for MANETs", IEEE
Transactions on Mobile Computing, Vol. 14, No. 10,
pp. 2101–2155, Oct. 2015.
[30] J. Duan, D. Gao, D. Yang, C.H. Foh, and H.H Chen,
“An Energy-Aware Trust Derivation Scheme With
Game Theoretic Approach in Wireless Sensor
Networks for IoT Applications”, IEEE Internet of
Things Journal, Vol. 1, No. 1, 2014.
[31] K. Drira, H. Seba, H. Kheddouci, “ECGK: An
efficient clustering scheme for group key
management in MANETs”, Computer
Communications, Vol.33, pp.1094–1107, 2010.
[32] T. Issariyakul and E. Hossain, Introduction to
Network Simulator NS2. New York, NY, USA,
Springer, 2011.